Limited POSEIDON

Ernesto Carrella

8/4/2020

Intro

Poseidon in Indonesia

  • Approach:
    • Take catch-at-length data
    • Feed it into length-based assessment techniques
    • Estimate population/carrying capacity
    • Run realistic POSEIDON over it
  • Advantages:
    • Best available science
    • Realistic
    • Can focus on details

Disadvantage

  • Opaque:
    • How data affects model
    • What is important
  • Slow:
    • Changes in assumptions/data require starting from scratch
  • Uncertainty poorly defined

A possible workaround

  • Keep your model simple
  • Isolate parameter/state space by rejection sampling
  • Focus less on:
    • The “right” parameters
    • The “right” distribution of parameters
  • Focus more on:
    • a large ensemble of valid histories

How rejection sampling works

Run the model “many” times

Start filtering

Keep filtering (SPR)

Keep filtering (Landings 2)

Keep filtering (SPR 2)

Simplified 712

Simplified model

  • 2 species (Lutjanus malabaricus + Pristipomoides multidens)
  • 10x10 simple map
  • 2 fleets (0-9; 10+)

Things we don’t know:

  • Initial population
  • Carrying capacity
  • Current depletion and the like

Things we know:

  • Last year landings
  • SPR
  • data we collected

Approach:

  • Start the fishery “ab novo”
  • Repeatedly:
    • Randomize all parameters \(K,L_{\inf},\dots\)
    • Run simplified POSEIDON for “many” years (up to 40)
      • Free entry
    • Accept simulation if model output are the same as what we observe in data
  • Collect

Why bother?

  • Benefits:
    • No need for population reconstruction
    • Clear uncertainty propagation
    • Clear value of information
  • Cost:
    • Model needs to be fast

Results

  • Ran the model for about 50,000 times
  • Accepted about 1.5% of them
  • This is their story….

Weighing evidence

Landings only scenario

  • Imagine we only knew what this year landings (no length data)
    • 5000t to 10000t for malabaricus
    • 1000t to 3000t for multidens

Knowing SPR

  • How does it change if I also add as a condition
    • OBSERVED SPR of malabaricus must be between 0.2% and 20% (measures from 2018 and 2016 but adding a bit of pessimism) respectively

Knowing SPR of other species

  • How does knowing that the SPR of the other species is between 10 and 20 help?

Landings only (2)

Knowing SPR (2)

Knowing SPR of other species (2)

Basic results

  • Once SPR is properly contextualized within the history of the fishery; these are actually green shoots
  • Measuring the SPR of another species can increase our expected estimate of depletion for this species

Additional knowledge

  1. I know that both boat populations are still at it
  2. I know that small boats are more numerous
  3. I know that malabaricus SPR has dropped dramatically?

Additional Knowledge 2

Additional Knowledge 3

Policy

Policy attempts

We don’t want just to assess, we want to know policy effects:

  • no policy
  • put a 100 day limit to 10GT only
  • put a 100 day limit to all

Data-limited

  • Works as before but:
    • apply same policy to all accepted runs
  • Positive:
    • study the average effect
    • understanding the uncertainty
  • Negative:
    • to make them comparable, speak only in relative terms

Results 1

Results Biomass

Results Landings

Results Cashflow

718

  • 3 Species
    • Lutjanus malabaricus
    • Atrobucca brevis
    • Lethrinus laticaudis
  • 3 Fleets
    • Large Longliners
    • Large Dropliners
    • Large Gillnetters

718 - Posteriors

718 - Policy

718 What’s missing

  • Price increase
  • Keep probolinggo boats very far
  • Deal with seasonality of dropliners